Repository navigation
Conversation
…ady say Each removed sentence is said again by every prompt that loads the skill, or by another skill that prompt loads. Nothing moves and no prompt changes. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
🧙 Wizard CIRun the Wizard CI and test your changes against wizard-workbench example apps by replying with a GitHub comment using one of the following commands: Test all apps:
Test all apps in a directory:
Test an individual app:
Show more apps
Test against a wizard branch:
Add Results will be posted here when complete. |
…drop two embedded goals Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
| Decide which custom events are worth capturing, then instrument them — in one | ||
| pass, reading each file once. |
There was a problem hiding this comment.
A goal, and the prompt states it. Superseded by context/agents/integration-v2/capture.md:17-19:
Decide which events are worth capturing in this app, then instrument them in the same pass — read each file once, choose the events, and add the capture calls while the file is already open.
| `lower_snake_case` names and useful properties. Edit each file while it is already | ||
| open. |
There was a problem hiding this comment.
Only "Edit each file while it is already open." is removed. Superseded by context/agents/integration-v2/capture.md:18-19:
read each file once, choose the events, and add the capture calls while the file is already open.
| in prose, without the marker, is dropped. Also record the dashboard URL in your | ||
| handoff so the report step can link it. |
There was a problem hiding this comment.
Only "Also record the dashboard URL in your handoff so the report step can link it." is removed. Superseded by context/agents/integration-v2/dashboard.md:25-26:
A dashboard exists with a handful of insights on the captured events, and you hand off its URL for the report to link.
| add the handler there. Do not read through the whole app or wrap individual | ||
| components or routes by hand. |
There was a problem hiding this comment.
Only "Do not read through the whole app or wrap individual components or routes by hand." is removed.
context/agents/integration-v2/error-tracking.md:40-42:
You did not install anything, run a build, lint, or tests (the review task verifies after you), search outside the project, or read through the whole app or hand-wrap individual components or routes.
context/agents/error-tracking/capture-exceptions.md:114-116:
You installed nothing but the server library your capture imports, and did not run tests, search outside the project, or read through the whole app or hand-wrap individual components or routes.
| Find the auth flow first: login and signup handlers, session callbacks. If the | ||
| app has no concept of a user, there is nothing to identify — report that and stop. |
There was a problem hiding this comment.
Only "If the app has no concept of a user, there is nothing to identify — report that and stop." is removed.
context/agents/integration-v2/identify.md:64:
If the app has no auth or user concept, say so and stop.
context/agents/support/identify.md:38:
If the app has no login, say so and stop.
| poetry, else pip; `composer.lock` → composer; `Gemfile.lock` → bundle). Add the | ||
| client library, plus the server library if the app sends events server-side. | ||
|
|
||
| Read the manifest first. If the SDK is already installed, leave it and say so. |
There was a problem hiding this comment.
Only "If the SDK is already installed, leave it and say so." is removed. All five prompts that load this skill say it.
context/agents/integration-v2/install.md:31:
If the SDK is already installed, leave it and say so.
context/agents/error-tracking/install.md:16-17, context/agents/replay-vision/install.md:16-17:
If it's already installed, leave it alone and say so in your handoff.
context/agents/feature-flags/install.md:19:
Leave an SDK that is already there alone and say so.
context/agents/support/install.md:16-17:
If it's already there, leave it alone and say so in your handoff.
| # Mirror the report into a PostHog notebook | ||
|
|
||
| Mirror the setup report into a shareable PostHog notebook, so the user has an | ||
| in-app copy to link and comment on. No report file is written to the project. |
There was a problem hiding this comment.
Only "No report file is written to the project." is removed. All four report prompts that load this skill say it, in the same words:
Do not write a report file.
context/agents/integration-v2/report.md:23, context/agents/feature-flags/report.md:21, context/agents/metrics/report.md:35, context/agents/support/report.md:20.
| Compose the full setup report as markdown, then publish it in **one** | ||
| `publish_handoff` call. That call is how the report reaches the user's wizard | ||
| session. Do not write a report file. The same call mirrors the markdown into a | ||
| shareable PostHog notebook. The notebook skill says what its result means. |
There was a problem hiding this comment.
A goal, and the prompt states it. Superseded by context/agents/integration-v2/report.md:21-24:
Then publish it with a single
publish_handoffcall. The same call mirrors the markdown into a shareable PostHog notebook. The notebook skill says what the result means and covers older wizards. Do not write a report file. The handoff call and the notebook are how the report reaches the user.
| Do not fall back to writing a file. Then follow the notebook skill for what | ||
| the result means. The notebook is best effort and never fails this step. | ||
| Record its outcome in your final message. Do not call `publish_handoff` again. |
There was a problem hiding this comment.
Three sentences removed, each said in context/agents/integration-v2/report.md.
| Removed | Prompt line |
|---|---|
| Do not fall back to writing a file. | :23 "Do not write a report file." |
| Then follow the notebook skill for what the result means. | :22-23 "The notebook skill says what the result means and covers older wizards." |
| The notebook is best effort and never fails this step. | :51-53 "The notebook mirror is best effort […] A missing notebook never fails this task." |
| 5. **Change no project code.** This step connects external data. It edits | ||
| nothing in the app. |
There was a problem hiding this comment.
Superseded by context/agents/integration-v2/warehouse.md:30:
Touch no project code. This step connects data; it does not edit the app.
The two tenets after it are renumbered, with no other change.
…t in the same words Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Removes 24 sentences from ten
integration-v2step skills. Each is a goal the task prompt already states, or a rule every prompt that loads the skill repeats, so the task agent is told the same things. Users see nothing change.What's new
integration-v2-report. Loses its intro and the closing publish reminders.integration-v2-build. Loses the test-suite ban and the done and failed rule.integration-v2-capture,-error-tracking-step,-identify. Lose two sentences each.integration-v2-init,-install,-dashboard,-warehouse,-notebook. Lose one each.Merge danger. Prose only, fetched by wizard task agents from the latest release. A revert restores the sentences. Blast radius: the task agents of the default, error-tracking, feature-flags, metrics, replay-vision and support flows.
How it flows
Nothing flows differently. No prompt, config, skill id or variant id changes.
The rule
A sentence went only when a line in every prompt that loads the skill says it. Each removal has an inline comment quoting that line.
Details and verification
Checks, against a build of
mainatff49233:SKILL.md, each reference file) was rebuilt for 49 tasks across every framework variant, 634 texts. 24 sentences are gone, the ones commented on. 0 are new. 0 reference files changed.npm test: 30 files, 339 tests pass.Kept, though something else repeats them:
dashboard's "create the insights anyway", which theinsightskill also says.install, parts ofinitandidentify.build's "it matches the codebase".🤖 Generated with Claude Code